HWY101 accident..

[Az_Rael]

Those of you who travel that portion of the 101 would know better but from pictures it appears this fatal event happened in a location where a manual steering correction would have had to be very rapid. While I have not purchased EAP and have only used it on service loaners, I do notice that when I assume control of the steering it's not exactly a silky smooth handoff. I also have read here about people stating that EAP can make some very sudden and unpredictable steering maneuvers. While I like to think my reflexes are pretty good and I definitely don't trust EAP to be infallible, I wonder if I would have been able to correct the steering in time as it veered into that abutment. I'd like to hear from those who travel that route, not the people who are being judgmental but have limited or no experience driving this route.

You might post this over in the big thread in the X forum. Several locals who commute that route daily are following that one and can maybe answer.

 

[Driver Dave]

For them, its ALWAYS never Tesla's fault.

Interested, how could something like this be Tesla’s fault? I really don’t see anything remotely close to it but am curious what you see.

 

[Doug_G]

People are emotional about a very important topic. Obviously I don't want to get killed, on the flip side, people who use autopilot without abuse, don't want their favorite luxury item taken away, because other people don't follow the instructions.

Energetic discussion is fine as far as it relates to autopilot. When people start making it personal then they need to cool it.

 

[techmaven]

First of all, there is a difference between the hitting the barrier and killing the driver. The killing the driver part has significant factor in that the attenuation barrier was not reset. A Prius slammed into the attenuation barrier which was properly set up more than a week earlier at 70+mph and the driver walked away.

Second, the human driver with level 2 tech is always, always, always in charge. Once the legal and technological responsibility shifts, there will be a lot of press and discussion about that. Until then, the human driver is responsible. The ADAS system here is just to provide assistance and the driver is to maintain situational awareness. Plenty of people get into accidents and die on the roads without ADAS systems. They fall asleep. They get distracted using their phones, fixating their stare on something else, reading a book, or putting on makeup. Accidents and fatalities happen even without ADAS systems on a pretty regular basis.

Third, there are a slew of circumstances that caused this accident. The poorly marked lanes and no cross-hatching on the split. The way the pavement with the asphalt and concrete is laid out can trick the ADAS system. The time of day with the sun can make visibility even worse. Putting all the blame on the Autopilot system is incorrect on many levels. But it is also correct that the Autopilot system drove Mr. Huang into the concrete barrier.

With ~5 seconds to notice the concrete barrier and realize that the car was drifting out of lane, the driver was clearly not paying sufficient attention. And the earlier warnings about Autopilot provides some indication that it wasn't the first time during this drive that the driver wasn't paying sufficient attention.

Also, I don't understand why the family pointed out that the driver had issues with Autopilot before. That indicates a heightened awareness that the Autopilot system needs to be supervised, especially if this location was specifically pointed out. It makes even less sense that this driver would leave Autopilot unattended if he had difficulties at this location with this technology.

 

[drewg123]

I'm kind of in the middle on all this. I think the blame is shared. When AP does something unexpectedly crazy, even if your hands are on the wheel, it can be difficult to react. However, if the driver had really reported problems at this interchange in the past, then he should have been watching it especially closely, and, IMHO, deserves the lions share of the blame. What I really don't like is how Tesla is framing things to put all the blame on the driver via the carefully worded statements about having AP2 nags at some point in the drive.




I have an intersection near my house (on surface streets) that is complicated and poorly marked, and has the same sort of "not a lane" extra lane issue as the 101/85 interchange (which I used to live ~2 miles from). After the latest 2018.10.4 update, when AP2 on surface streets became much more usable, I gave this intersection a try at a very low speed. The car did what I'm afraid may have happened in this accident. It was following the right turn / exit path, and then got confused, and abruptly steered *HARD* left into the "not a lane" lane. I was going slow (25mph) with no traffic nearby, and was watching it like a hawk, so its no big deal. But I could imagine being startled at highway speeds if I wasn't expecting it.

 

[mblakele]

With AutoPilot and similar technologies in play, road accidents are starting to look more and more like airplane accidents. I'm no expert on those, but I've read analyses of a few of them. The notion of "fault" tends to be a little different, because crashes aren't usually caused by any single pilot error or system failure. Instead there's often a chain of failures and errors leading to the crash, without each of which the crash wouldn't have happened. After figuring that out, the folks involved try to address as many of them as possible. The goal is that there'll never be another crash with exactly the same sequence of failures and errors. For example in this case the analysis might recommend better road markings and high-priority replacement of crash barriers in this location, plus changes to AP to mitigate driver inattention and to improve AutoSteer and AEB.

However another lesson from flight safety is to avoid public speculation until all the evidence is available. We may not know everything yet.

 

[lasdoug]

You might post this over in the big thread in the X forum. Several locals who commute that route daily are following that one and can maybe answer.

You might post this over in the big thread in the X forum. Several locals who commute that route daily are following that one and can maybe answer.

First of all, there is a difference between the hitting the barrier and killing the driver. The killing the driver part has significant factor in that the attenuation barrier was not reset. A Prius slammed into the attenuation barrier which was properly set up more than a week earlier at 70+mph and the driver walked away.

Second, the human driver with level 2 tech is always, always, always in charge. Once the legal and technological responsibility shifts, there will be a lot of press and discussion about that. Until then, the human driver is responsible. The ADAS system here is just to provide assistance and the driver is to maintain situational awareness. Plenty of people get into accidents and die on the roads without ADAS systems. They fall asleep. They get distracted using their phones, fixating their stare on something else, reading a book, or putting on makeup. Accidents and fatalities happen even without ADAS systems on a pretty regular basis.

Third, there are a slew of circumstances that caused this accident. The poorly marked lanes and no cross-hatching on the split. The way the pavement with the asphalt and concrete is laid out can trick the ADAS system. The time of day with the sun can make visibility even worse. Putting all the blame on the Autopilot system is incorrect on many levels. But it is also correct that the Autopilot system drove Mr. Huang into the concrete barrier.

With ~5 seconds to notice the concrete barrier and realize that the car was drifting out of lane, the driver was clearly not paying sufficient attention. And the earlier warnings about Autopilot provides some indication that it wasn't the first time during this drive that the driver wasn't paying sufficient attention.

Also, I don't understand why the family pointed out that the driver had issues with Autopilot before. That indicates a heightened awareness that the Autopilot system needs to be supervised, especially if this location was specifically pointed out. It makes even less sense that this driver would leave Autopilot unattended if he had difficulties at this location with this technology.

____________________________________________________________________________________________________________
If what I’m reading regarding the accident is factual (Autopilot Activated – Several Visual and One Audible Hand Warnings – Hands were not on the wheel 6 seconds prior to accident – Driver had 5 seconds unobstructed view of concrete barrier and No Action was taken) then I agree with (Techmaven) 100%.

Being a Retired 767 Captain, I would like to chime in, regarding “AUTOPILOT”.

Having flown with one of the BEST autopilots in service, our favorite saying in the cockpit
Regarding the Autopilot was “WHATS IT DOING NOW” Meaning, the autopilot is doing
something totally unexpected, which is why we always maintained a constant vigil in the cockpit.

We had the luxury of a little more time in the air to correct a situation unlike a Model X traveling 70 mph on the highway. Also you can bet your AS_ if we heard a CHIME or saw a VISUAL WARNING we were correcting the situation, NOW!!

Having never driven a Tesla (Had a model 3 on order for 2 years up till today),
I think the word “AUTOPILOT” in the Tesla Lingo, along with a drivers daily experience using a STABLE, WELL PERFORMING AUTOPILOT might lead to complacency behind the wheel.

PROBLEM IS- Its only takes one Bad road design or crumbling infrastructure and being complacent behind the wheel to end in a tragic accident. I can only imagine how much worse this accident could have been if nearby drivers were also killed in this accident.

When I see YouTube videos of some people driving their Tesla’s with BOTH FEET out the driver’s window, I want to scream because it only takes a FEW to kill a great innovation like AUTOPILOT, for the rest of us.

If the Government says, “OK that’s it, 2 deaths in 2 weeks” (No matter who’s at fault), no more vehicle automation, then a few just ruined a great tool for the rest of us.

Just my Thoughts.

 

[Doug_G]

Think about it this way: autopilot follows painted lines. That is pretty much all it knows*. The only thing keeping your vehicle on the road is a camera following lines painted on the road. That's a fairly feeble barrier against a one ton vehicle crashing into the ditch, especially if the markings are faded or confused. Therefore it is absolutely necessary to pay close attention when driving on autopilot.

Sure, 99.9% of the time the car follows the lines just fine. You really need to be alert for those occasions where it doesn't. Anyone who has done even a modest road trip has experienced this.

(*Okay I've noticed that recent versions also slow down when they see flashing lights. But mostly it just follow paint.)

 

[mariner22]

I don't have any experience with autopilot (I'm still waiting for a Model 3) but isn't it supposed to function as a higher level cruise control? Many cars have cruise control - but it is obviously the drivers fault if you set the car to cruise at 60 mph and then run into a car going much slower.

 

[croman]

Think about it this way: autopilot follows painted lines. That is pretty much all it knows*. The only thing keeping your vehicle on the road is a camera following lines painted on the road. That's a fairly feeble barrier against a one ton vehicle crashing into the ditch, especially if the markings are faded or confused. Therefore it is absolutely necessary to pay close attention when driving on autopilot.

Sure, 99.9% of the time the car follows the lines just fine. You really need to be alert for those occasions where it doesn't. Anyone who has done even a modest road trip has experienced this.

(*Okay I've noticed that recent versions also slow down when they see flashing lights. But mostly it just follow paint.)

That is a very simplistic view. AP1 and 2 consider additional data including road edge, curbs, and for AP1 it recognizes speed limits. AP2 has been indicated to also detect hills (both up and down), construction zones, and, since 2018.10.4, road wear (i.e. tire wear on roads).

I've driven over 20k miles now on AP2. I use it all the time and I love it but I'm aware that the system can fail me. I don't text and drive (AP or not) but I do think the system is not operating as designed because I've had 5-7 incidents over those 20k miles where AS tried to kill me. They mostly occurred in earlier versions but it was not great at following lane lines (AP2) and even 2018.12 will make very dangerous decisions based on an imperfect reading of the lane lines (though it clearly thinks its doing its job).

 

[croman]

I don't have any experience with autopilot (I'm still waiting for a Model 3) but isn't it supposed to function as a higher level cruise control? Many cars have cruise control - but it is obviously the drivers fault if you set the car to cruise at 60 mph and then run into a car going much slower.

No, that is dumb cruise. Tesla uses TACC (traffic aware cruise control). It sensor fuses radar and camera(s) (and sorta uses ultrasonics at speed but only the side ones for "collision avoidance") to detect objects around it.

Theoretically, the crash attenuator should've been detected by AP and the car should've engaged AEB when the crash was deemed unavoidable but the severity of the crash indicates that no AEB engaged.

If my car rammed the car ahead, clearly I wasn't paying enough attention but it is equally true that AP could fail and ram you into a car in an instant before you could react. In one situation, it could be my fault, in the latter, it is AP's fault.

 

[WannabeOwner]

I won't pay the money for AP or EAP software until I am convinced it is so good that even if I lose focus, it will prevent a senseless catastrophe like the 101 accident

Meantime if you lose focus, for any reason, you have no backup? I'm the reverse: much happier to have AP watching my back ...

 

[Doug_G]

That is a very simplistic view. AP1 and 2 consider additional data including road edge, curbs, and for AP1 it recognizes speed limits. AP2 has been indicated to also detect hills (both up and down), construction zones, and, since 2018.10.4, road wear (i.e. tire wear on roads).

I've driven over 20k miles now on AP2. I use it all the time and I love it but I'm aware that the system can fail me. I don't text and drive (AP or not) but I do think the system is not operating as designed because I've had 5-7 incidents over those 20k miles where AS tried to kill me. They mostly occurred in earlier versions but it was not great at following lane lines (AP2) and even 2018.12 will make very dangerous decisions based on an imperfect reading of the lane lines (though it clearly thinks its doing its job).

I don't trust it next to unpainted curbs. To be fair it's been a while since I've tried it.

Still, while AP has been clearly getting more sophisticated, no one should assume it has any "intelligence" at this point. If you assume it's following simple rules you won't be surprised when it does something dumb.

 

[Doug_G]

Theoretically, the crash attenuator should've been detected by AP and the car should've engaged AEB when the crash was deemed unavoidable but the severity of the crash indicates that no AEB engaged.

Practically speaking, the radar would not have detected the (broken) crash attenuator because it is a stationary object. The radar can't reliably detect stationary objects because of the huge amount of ground clutter from other stationary objects. If it tried to do this the car would be doing panic stops all the time.

As for detecting it optically, it's pretty clear the current system does not have anywhere near that level of sophistication.

 

[_jmk]

for detecting it optically, it's pretty clear the current system does not have anywhere near that level of sophistication.

The original question was, can it ever have? There will always be things it cannot classify. Certainly human brain can, but it is entirely different machine we do not yet understand.

 

[R.S]

To add to this even with this fatal accident the Tesla system is proven to be 3,7 times more safe than a human driver.

Given that I am 3,7 Times more likely to survive in a Tesla with AP if I follow the rules. If you don’t you should not be surprised if you get into accidents and dangerous situations.

It’s for me therefore a very easy decision that I will use always Tesla AP with Full attention to the road as this will help me to avoid any kind of accidents and keeps me more safe than other drivers.

You are 3.7 times less likely to die in a Tesla with AP hardware, than in the average car.

But in 2013 the Insurance Institute for Highway Safety did a statistic over cars of the model years 2009-2012 and the Honda Odyssey had 0 fatalities, as well as 8 other cars in those four years. And the Honda Odyssey was sold more often than AP HW Teslas.

Why is that? Maybe because Odysseys are built safe, maybe because the drivers are parents that drive more carefully. Who knows, but is a 2009 Odyssey an upgrade to a 2017 Tesla when it comes to safety? Probably not, but who can really say?

Other reasons why the fatality rate in a Tesla with AP HW is lower than in the average car:
-Tesla customers are affluent enough to do regular maintenance, like getting new tires, checking the brakes, alignment etc.
-The cars are pretty new, so not many neglected cars
-The people driving Teslas are usually not teenagers and right now Teslas aren't cars very old people drive, both the highest risk drivers
-Probably not many substance abusers can afford a Tesla.
-Teslas are inherently super safe cars, with extremely good crash test results.

So saying, that driving with AP on reduces your fatality rate while driving 3.7 times, is probably not right.

 

[jelloslug]

I don't trust it next to unpainted curbs. To be fair it's been a while since I've tried it.

Still, while AP has been clearly getting more sophisticated, no one should assume it has any "intelligence" at this point. If you assume it's following simple rules you won't be surprised when it does something dumb.

I have found that unpainted curbs are not very good as a lane marker either.

 

[croman]

-The people driving Teslas are usually not teenagers and right now Teslas aren't cars very old people drive, both the highest risk drivers
-Probably not many substance abusers can afford a Tesla.

I don't disagree with your overall point but your generalizations get you into trouble. I know of more 65+ Tesla owners than any other group.

Further, affluence and drug use are not related. Poor people abuse drugs and so do rich people. Rich people can afford different drugs, but the incidences of abuse are fairly similar. Its factually incorrect to say substance abusers cannot afford Teslas. There are thousands of Teslas in the North Shore and I read the news about several that are arrested for DUI (substance abuse). I'm sure a lot more are just not arrested or are on other substances that are not readily detectable during traffic stops.

 

[R.S]

I don't disagree with your overall point but your generalizations get you into trouble. I know of more 65+ Tesla owners than any other group.

Further, affluence and drug use are not related. Poor people abuse drugs and so do rich people. Rich people can afford different drugs, but the incidences of abuse are fairly similar. Its factually incorrect to say substance abusers cannot afford Teslas. There are thousands of Teslas in the North Shore and I read the news about several that are arrested for DUI (substance abuse). I'm sure a lot more are just not arrested or are on other substances that are not readily detectable during traffic stops.

Sure those were assumptions, but 65 isn’t really as much a problem. It’s 75+, where it gets tough. The safest years are actually 30-74.

Statistically the cars you are going to have the highest rate of accidents in is your first and your last. Both will probably not be a Model S and X right now.

But in general I just wanted to give some reasons why fatalities might be lower. Not all of them might have to be true, but I think it’s safe to say that even w/o AP the cars would have much lower fatality rates, than the average car.

Sure, maybe somewhere out there there is a 16 year old guy doing drunk street races in his Model X with totally worn down brakes and tires, busted headlights and halve the airbags torn out, but probably not.

 

[avoigt]

You are 3.7 times less likely to die in a Tesla with AP hardware, than in the average car.

But in 2013 the Insurance Institute for Highway Safety did a statistic over cars of the model years 2009-2012 and the Honda Odyssey had 0 fatalities, as well as 8 other cars in those four years. And the Honda Odyssey was sold more often than AP HW Teslas.

Why is that? Maybe because Odysseys are built safe, maybe because the drivers are parents that drive more carefully. Who knows, but is a 2009 Odyssey an upgrade to a 2017 Tesla when it comes to safety? Probably not, but who can really say?

Other reasons why the fatality rate in a Tesla with AP HW is lower than in the average car:
-Tesla customers are affluent enough to do regular maintenance, like getting new tires, checking the brakes, alignment etc.
-The cars are pretty new, so not many neglected cars
-The people driving Teslas are usually not teenagers and right now Teslas aren't cars very old people drive, both the highest risk drivers
-Probably not many substance abusers can afford a Tesla.
-Teslas are inherently super safe cars, with extremely good crash test results.

So saying, that driving with AP on reduces your fatality rate while driving 3.7 times, is probably not right.

Well, numbers don't lie. Either they have less fatal accidents or they don't. I appreciate your long response but there is not dispute about the 3.7 time more safer fact. It is like it is. I could go deeper into statistics but what I read out of you mail is that you are not very familiar with it. No offense because most people are not. However allow me to say that all your arguments listed don't apply given the statistics. I learned it at University and believe me there is no point for you to make about this number.

Happy to hear that Honda 5 years ago has a safe car back in 2009 - 2012 but what does this have to do with Tesla and AP today?

Take it as that and if you make the right conclusions it may safe your life one day as well..... you never know.